AoI Aware Channel Scheduling for V2N Status Reporting Based on Deep Reinforcement Learning

In this paper, we focus on time-division-multiple-access based channel scheduling for vehicular-to-network (V2N) status reporting service over fast fading channels, where the base station (BS) cares about the age of information (AoI) regarding vehicles' status. To provide min-max fair long-term AoI performance for all vehicular users, we propose a deep reinforcement learning based channel scheduling algorithm, leading to a mapping from possible network situation to user selection. First, the problem is formulated as an infinite-horizon Markov decision process with infinite state and finite action spaces, where the reward is designed as the minus value of the maximum AoI of all vehicular users. Then, a deep Q-network is applied to generate an efficient channel scheduling policy. According to our simulation, the proposed algorithm has better performance on maximum AoI and AoI outage probability comparing with the considered four benchmark approaches.